Influence of the number of vulnerability classes in determining morphometric susceptibility to flooding

Abstract Watershed morphometry is an important tool to diagnose susceptibility to flooding. This diagnosis can guide the planning and implementation of mitigation measures to avoid flooding damage. The objective of this work was to study flooding susceptibility in Espírito Santo Stream Basin (ESSB), located in the municipality of Juiz de Fora, Minas Gerais, Brazil, based on its morphometric characterization using two, three and five classes of vulnerability. In order to accomplish this, the basin was divided considering two levels of detail: in the first, less-detailed level, three sub-basins were generated; and in the second, more-detailed level, 65 watersheds were generated. Morphometric susceptibility to flooding was evaluated in these study units and the results were compared using the K-means and Fuzzy C-means clustering techniques. The results showed that the number of classes adopted affects the result of classification vulnerability, suggesting that cluster validation criteria can be used to delimit the choice of the number of classes to be adopted. Groups formed only by morphometric similarities differ from those obtained by the methodology based on classification of areas susceptible to flooding, since the methodology for susceptibility analysis transforms the numerical values of each morphometric parameter in a class of susceptibility.

摘要：流域形态计量学（Watershed morphometry）是诊断洪水易感性的重要工具，该诊断结果可指导减灾措施的规划与实施，规避洪涝灾害损失。本研究以巴西米纳斯吉拉斯州茹伊斯迪福拉市境内的圣埃斯皮里图托溪流域（Espírito Santo Stream Basin, ESSB）为研究对象，基于其形态计量特征，采用2类、3类及5类易损性等级开展洪水易感性分析。为达成上述研究目标，本研究基于两级细节尺度对该流域进行划分：第一级为低细节尺度，生成3个子流域；第二级为高细节尺度，生成65个小流域。针对上述研究单元开展洪水易感性的形态计量评估，并采用K-means聚类与模糊C均值聚类（Fuzzy C-means clustering）两种技术对结果进行对比分析。研究结果表明，所采用的易损性等级数量会对易损性分类结果产生影响，提示可借助聚类验证准则确定适宜的等级数量选取方案。仅基于形态计量相似性形成的聚类组，与基于洪水易感性区域分类方法得到的聚类组存在差异，原因在于易感性分析方法会将各形态计量参数的数值转换为对应的易损性等级。